Intelligent three-dimensional model design method and system for spare and accessory parts

Abstract

The invention relates to the technical field of intelligent three-dimensional model design, in particular to an intelligent three-dimensional model design method and system for spare and accessory parts. The method comprises the following steps of performing multi-dimensional perception data acquisition on the spare and accessory parts to obtain an original sensor data stream; performing associated feature parameter mining on the original sensor data stream to obtain a primary feature parameter set; performing multi-dimensional feature fusion construction on the primary feature parameter set to obtain a behavior feature vector; historical fault data of the spare and accessory parts are obtained; performing multi-modal failure prediction model training according to the historical fault data of the spare and accessory parts to obtain a multi-modal failure prediction model; and performing future failure probability deduction by using the multi-modal failure prediction model, and performing time sequence failure risk accumulation to obtain failure probability distribution. According to the invention, the design of spare and accessory parts can better adapt to the actual operation environment, so that the reliability is improved, the cost is reduced, and the efficiency is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent three-dimensional model design, and in particular to an intelligent three-dimensional model design method and system for spare parts. Background Art

[0002] The design method of spare parts has changed from traditional experience-driven to data-driven. Although the traditional CAD / CAM method has achieved the digitization of the model, it still relies on the experience and knowledge of the designer. Therefore, some emerging technologies have begun to be applied to the field of spare parts design, promoting the development of intelligent design methods. For example, by defining parameters and constraints, the model can automatically adjust the geometric shape according to the change of parameters, improving the flexibility and efficiency of the design. Using methods such as finite element analysis (FEA) and computational fluid dynamics (CFD), the performance of spare parts is virtually verified, reducing the cost of making and testing physical prototypes.

[0003] However, the existing intelligent 3D model design methods still have some limitations: for example, design redundancy or insufficiency, high cost, low efficiency, parametric modeling and topology optimization can improve design efficiency, but still require designers to manually set parameters and constraints, and it is difficult to fully consider the complexity of actual operating conditions. This leads to design redundancy (over-design, increased material and manufacturing costs) or insufficiency (insufficient design strength, resulting in premature failure of parts). Existing methods are difficult to design and optimize in advance according to actual operating conditions and potential failure risks. Simulation analysis is mainly used to verify existing design solutions, and it is difficult to actively predict potential failure risks. Summary of the invention

[0004] Based on this, it is necessary to provide an intelligent three-dimensional model design method and system for spare parts to solve at least one of the above technical problems.

[0005] To achieve the above object, a method for intelligent three-dimensional model design of spare parts includes the following steps:

[0006] Step S1: acquiring multi-dimensional sensing data of spare parts to obtain original sensor data stream; mining associated feature parameters of the original sensor data stream to obtain a primary feature parameter set; and constructing multi-dimensional feature fusion on the primary feature parameter set to obtain a behavior feature vector;

[0007] Step S2: Obtain historical failure data of spare parts; train a multimodal failure prediction model based on the historical failure data of spare parts to obtain a multimodal failure prediction model; use the multimodal failure prediction model to deduce future failure probabilities and perform sequential failure risk accumulation to obtain a failure probability distribution;

[0008] Step S3: performing failure mechanism and design parameter correlation analysis on the failure probability distribution to obtain a failure-parameter correlation table; performing design parameter sensitivity evaluation on the failure-parameter correlation table to obtain parameter sensitivity evaluation data; performing multi-objective constraint construction and optimization based on the parameter sensitivity evaluation data to obtain optimized design constraints;

[0009] Step S4: assigning constraint-driven parameters according to the optimization design constraints to obtain a constraint assignment model; constructing a geometric morphology adaptive three-dimensional model according to the constraint assignment model to obtain an adaptive three-dimensional model;

[0010] Step S5: monitor the new model service status of the adaptive three-dimensional model, and generate a maintenance file to obtain a new model maintenance file; generate a service performance index according to the new model maintenance file to obtain a service performance index.

[0011] The present invention effectively extracts key features reflecting the running status of parts and components through multi-dimensional perception data acquisition, associated feature parameter mining and multi-dimensional feature fusion construction, forms behavioral feature vectors, provides high-quality data basis for subsequent failure prediction, avoids information loss and data redundancy, and improves the accuracy and efficiency of the prediction model. By integrating historical fault data, training multi-modal failure prediction models, and performing future failure probability deduction and temporal failure risk accumulation, accurate prediction of future failure risks of parts and components is achieved, which can not only predict failure probability, but also predict failure mode and time distribution, providing clear goals and directions for subsequent design optimization. Through the analysis of failure mechanism and design parameter association, design parameter sensitivity evaluation and multi-objective constraint construction and optimization, the failure prediction results are converted into operational design constraints, and the optimal design parameter combination is found through the multi-objective optimization algorithm, realizing adaptive design based on failure risk, thereby improving the reliability and life of parts and components. Through constraint-driven parameter assignment and geometric morphology adaptive three-dimensional model construction, rapid model generation based on optimized design constraints is realized, and a three-dimensional model that meets design requirements is automatically constructed. Through virtual performance evaluation and iterative optimization, the performance and reliability of the model are further improved, the design cycle is shortened, and the design efficiency is improved. By monitoring the service status of the new model, generating maintenance files and service performance indicators, a closed-loop feedback optimization mechanism is constructed, and the actual operation data is fed back to the design process, which can continuously verify and improve the design method and model, and achieve continuous optimization, so that the design of parts can better adapt to actual working conditions, further improve reliability, reduce costs and improve efficiency. Therefore, the present invention provides an intelligent three-dimensional model design method and system for spare parts, which realizes data-driven adaptive design and closed-loop feedback optimization by deeply integrating predictive maintenance data with the intelligent three-dimensional model design process, effectively solving the limitations of existing methods such as design redundancy or deficiencies, high cost, low efficiency, and difficulty in pre-design and optimization according to actual operating conditions and potential failure risks, so that the design of spare parts can better adapt to the actual operating environment in which they are located, thereby improving their reliability, reducing costs and improving efficiency.

[0012] Preferably, step S1 comprises the following steps:

[0013] Step S11: acquiring multi-dimensional sensing data of spare parts at key parts of the equipment to obtain original sensor data stream;

[0014] Step S12: performing heterogeneous data synchronization calibration on the original sensor data stream to obtain a synchronous calibration data set;

[0015] Step S13: performing noise interference identification and suppression on the synchronous calibration data set to obtain high signal-to-noise ratio data;

[0016] Step S14: mining associated feature parameters for high signal-to-noise ratio data to obtain a primary feature parameter set;

[0017] Step S15: construct a multi-dimensional feature fusion on the primary feature parameter set to obtain a behavior feature vector.

[0018] The present invention can comprehensively obtain the operating status information of spare parts by deploying multiple types of sensors at key parts of the equipment and synchronously collecting data, providing a rich data source for subsequent feature extraction and failure prediction, avoiding the information loss that may be caused by relying solely on single sensor data, thereby improving the accuracy and reliability of subsequent analysis. By performing time synchronization and dimension calibration on multi-source sensor data, the data differences between different sensors, such as different sampling frequencies, inconsistent dimension units, etc., are eliminated, ensuring the consistency and comparability of the data, laying a foundation for subsequent feature extraction and data fusion, and improving the efficiency and accuracy of data processing. By adopting a suitable denoising algorithm to suppress the noise of sensor data, the noise caused by environmental interference and sensor errors can be removed, the signal-to-noise ratio of the data can be improved, so that the extracted feature parameters can better reflect the actual operating status of spare parts, thereby improving the accuracy and reliability of subsequent failure prediction. By performing feature extraction on high signal-to-noise ratio data, the original time domain or frequency domain signal can be converted into more representative feature parameters, such as vibration spectrum characteristics, temperature change rate, etc. These feature parameters can more effectively reflect the operating status and potential failure modes of spare parts, and provide more effective data input for subsequent feature fusion and failure prediction. By fusing multiple primary characteristic parameters into a multi-dimensional behavioral characteristic vector, the operating status of spare parts can be described more comprehensively, avoiding the information loss that may be caused by relying on a single characteristic parameter, thereby improving the accuracy and reliability of subsequent failure prediction. At the same time, the standardization process eliminates the dimensional differences between different characteristic parameters, so that each characteristic parameter has the same weight in the behavioral characteristic vector, avoiding the influence of certain characteristic parameters being overly amplified or reduced.

[0019] Preferably, step S14 comprises the following steps:

[0020] Step S141: constructing a mechanical model according to the high signal-to-noise ratio data to obtain a mechanical model;

[0021] Step S142: performing stress history deduction based on high signal-to-noise ratio data and a mechanical model to obtain a stress time series;

[0022] Step S143: extracting the cyclic load spectrum of the stress time series to obtain a cyclic characteristic spectrum;

[0023] Step S144: Calculate fatigue damage parameters according to the cycle characteristic spectrum to obtain fatigue damage index;

[0024] Step S145: performing common feature integration on fatigue damage indicators to obtain a primary feature parameter set.

[0025] The present invention links the load, displacement and other data collected by the sensor with the stress state of the component by constructing a mechanical model, thereby providing a basis for subsequent fatigue analysis. The establishment of a mechanical model makes it possible to more accurately evaluate the stress distribution of components under actual working loads, avoiding the errors caused by relying solely on experience or simplified formulas for estimation, thereby improving the accuracy of fatigue damage prediction. By inputting high signal-to-noise ratio data into the mechanical model for simulation calculation, the stress time series of key parts of the component can be obtained. The series accurately reflects the stress change history of the component during actual operation, providing more refined and closer to reality data for fatigue damage analysis, avoiding the errors caused by using simplified load spectra, thereby improving the accuracy of fatigue damage prediction. By extracting the cyclic load spectrum of the stress time series, complex variable amplitude loads can be converted into a series of cyclic loads with different amplitudes and means, simplifying the calculation process of fatigue analysis, while retaining the key features of the load spectrum, providing more convenient and more effective data input for subsequent fatigue damage calculations. By combining the SN curve and cyclic characteristic spectrum of the material, the fatigue damage indicators of the parts, such as the fatigue damage accumulation or equivalent stress amplitude, can be quantitatively calculated. These indicators can more intuitively reflect the fatigue damage degree of the parts and provide an important reference for subsequent life prediction and design optimization. By integrating the fatigue damage indicators with the common characteristic parameters extracted from other sensor data, a more comprehensive characteristic parameter set can be constructed. This parameter set not only contains the fatigue damage information of the parts, but also other important operating status information, such as vibration, temperature, etc., so as to more comprehensively describe the status of the parts, provide richer information for subsequent failure prediction and health assessment, and improve the accuracy and reliability of the prediction.

[0026] Preferably, step S2 comprises the following steps:

[0027] Step S21: Acquire historical fault data of spare parts; integrate historical fault information of spare parts historical fault data to obtain standardized fault files;

[0028] Step S22: labeling the behavioral feature vector and the standardized fault file in association with the behavioral feature to obtain a fault labeling feature set;

[0029] Step S23: training a multimodal failure prediction model using the fault annotation feature set to obtain a multimodal failure prediction model;

[0030] Step S24: inputting the behavior feature vector data into the multi-modal failure prediction model to deduce the future failure probability and obtain the immediate failure probability;

[0031] Step S25: Accumulate the instantaneous failure probability in terms of sequential failure risk to obtain a failure probability distribution.

[0032] The present invention collects and integrates historical fault data, and performs pre-processing such as cleaning, deduplication, and standardization to construct a structured standardized fault file, which provides a high-quality data basis for subsequent fault association analysis and model training, avoids the influence of data noise on the analysis results, and improves the accuracy and reliability of subsequent steps. By temporally correlating and labeling the behavior feature vector with the standardized fault file, a fault labeling feature set is constructed, and the operating state characteristics of the parts and components are associated with their corresponding failure modes, providing necessary supervised learning data for training the failure prediction model, so that the model can learn the potential relationship between the operating state characteristics and the failure mode. The multimodal failure prediction model is trained using the fault labeling feature set, so that the model can predict the probability of different failure modes occurring in the future according to the operating state characteristics of the parts and components, and provide a decision-making basis for subsequent preventive maintenance and design optimization. The application of the GBDT model can effectively process high-dimensional features and nonlinear relationships, and improve the accuracy of prediction. By inputting the real-time behavior feature vector into the trained failure prediction model, the instantaneous probability of different failure modes occurring at the current moment of the parts and components can be obtained, which provides a basis for real-time monitoring and early warning, and can timely discover potential failure risks. By accumulating the instantaneous failure probability over time, we can obtain the probability distribution of different failure modes of components in the future, more comprehensively evaluate the future failure risk, avoid the one-sidedness caused by focusing only on instantaneous probability, and provide a more reliable basis for formulating more reasonable maintenance strategies and design optimization solutions.

[0033] Preferably, step S3 comprises the following steps:

[0034] Step S31: performing high-risk failure screening on the failure probability distribution to obtain a significant failure mode set;

[0035] Step S32: performing correlation analysis between failure mechanism and design parameters on the significant failure mode set to obtain a failure-parameter correlation table;

[0036] Step S33: performing design parameter sensitivity evaluation on the failure-parameter association table to obtain parameter sensitivity evaluation data;

[0037] Step S34: generating initial design constraints according to the parameter sensitivity evaluation data to obtain an initial design constraint set;

[0038] Step S35: performing multi-objective constraint optimization and fusion on the initial design constraint set to obtain optimized design constraints.

[0039] The present invention sets a probability threshold to screen out failure modes with a high probability of occurrence in the future, and forms a significant failure mode set, which helps to concentrate resources to focus on and optimize these high-risk failure modes, improve the efficiency and pertinence of design improvement, and avoid the same degree of analysis of all failure modes, thereby optimizing resource allocation. By analyzing the failure mechanism of the significant failure mode and associating it with the design parameters of the parts, a failure-parameter association table is constructed, which design parameters have a greater impact on the specific failure mode, and provides a direction and basis for subsequent parameter sensitivity evaluation and design optimization, avoiding blind modification of design parameters. By performing sensitivity evaluation on the design parameters in the failure-parameter association table, the degree of influence of each parameter on the failure probability is quantified, which helps to identify the key design parameters with the greatest impact on the failure, and provides more accurate guidance for subsequent design optimization, avoiding excessive adjustment of minor parameters. According to the parameter sensitivity evaluation data, the initial design constraints are formulated for the high-risk failure mode, providing the initial search direction and range for the subsequent design optimization, converting the prediction results of the failure probability into operable design constraints, and providing specific goals for model optimization. By optimizing and integrating the initial design constraint set through a multi-objective optimization algorithm, a set of optimal design constraints that meet multiple design goals (such as reducing the probability of multiple failure modes, meeting weight and cost constraints, etc.) can be obtained, thereby achieving a balance between multiple competing design goals, avoiding the loss of one thing while focusing on another due to single-objective optimization, and achieving overall performance improvement of components.

[0040] Preferably, step S32 includes the following steps:

[0041] Step S321: Perform preliminary knowledge-based parameter identification on the significant failure mode set to obtain a preliminary associated parameter list;

[0042] Step S322: quantifying the parameter impact driven by simulation on the preliminary associated parameter list to obtain parameter sensitivity data;

[0043] Step S323: experimentally verifying the parameter sensitivity data and calibrating the correlation relationship to obtain calibrated parameter sensitivity;

[0044] Step S324: Develop a failure mechanism association rule set based on the preliminary association parameter list and the calibrated parameter sensitivity.

[0045] The present invention preliminarily identifies potential design parameters related to significant failure modes based on existing knowledge bases and literature retrieval, forms a preliminary association parameter list, provides direction for subsequent simulation analysis and experimental verification, avoids blindly analyzing all design parameters, improves analysis efficiency, and ensures that the focus of analysis is on the parameters that are most likely to affect the failure mode. Through simulation analysis, the influence of each parameter in the preliminary association parameter list on the failure mode is quantified, and parameter sensitivity data is obtained, which provides basic data for subsequent experimental verification and association relationship calibration, and can preliminarily screen out key parameters that have a greater impact on the failure mode before experimental verification, thereby reducing the workload and cost of experimental verification. Through experimental verification and calibration, possible errors in simulation analysis are corrected, more accurate and reliable parameter sensitivity data are obtained, the accuracy and reliability of the failure mechanism association rule set are improved, and the design optimization based on the rule set is more effective. By combining the preliminary association parameter list with the calibrated parameter sensitivity data, a failure mechanism association rule set is constructed, which clearly expresses the association relationship between design parameters and failure modes in the form of IF-THEN, forming a knowledge base that can be used to guide component design optimization, and providing a basis for automated and intelligent design optimization.

[0046] Preferably, step S33 includes the following steps:

[0047] Step S331: performing local sensitivity calculation on the failure-parameter association table to obtain local sensitivity data;

[0048] Step S332: performing a global sensitivity analysis according to the failure-parameter association table and the failure probability distribution to obtain a global sensitivity index;

[0049] Step S333: performing simulation-assisted sensitivity verification according to the failure-parameter association table to obtain simulation verification sensitivity data;

[0050] Step S334: Generate parameter sensitivity evaluation data for the local sensitivity data, the global sensitivity index, and the simulation verification sensitivity data to obtain parameter sensitivity evaluation data.

[0051] The present invention quantifies the influence of each design parameter on the failure probability near the reference value by calculating the local sensitivity, provides an understanding of the immediate influence of parameter changes, provides basic data for subsequent global sensitivity analysis and design optimization, and helps to identify which parameters have the greatest influence on the failure probability near the current design point. Through global sensitivity analysis, the influence of parameter changes in the entire value range on the failure probability is considered, and a more comprehensive sensitivity index, such as the Sobol index, is obtained, which can more accurately identify the key parameters that have the greatest influence on the failure probability, avoid the one-sidedness brought by local sensitivity analysis, and provide more reliable guidance for design optimization. Through simulation-assisted sensitivity verification, the results of local sensitivity and global sensitivity analysis can be further verified and calibrated, the accuracy and reliability of sensitivity evaluation can be improved, and the influence mechanism of parameter changes on failure modes can be understood. By comprehensively considering local sensitivity data, global sensitivity indexes and simulation verification sensitivity data, and adopting weighted average and other methods for fusion, more comprehensive and reliable parameter sensitivity evaluation data are obtained, which integrates the results of multiple analysis methods, improves the accuracy and robustness of the evaluation results, and provides a more reliable basis for the subsequent initial design constraint generation.

[0052] Preferably, step S4 comprises the following steps:

[0053] Step S41: applying the optimization design constraint to the pre-built initial parameterized model, performing constraint-driven parameter assignment, and obtaining a constraint assignment model;

[0054] Step S42: performing geometric shape adaptive generation according to the constraint assignment model to obtain a three-dimensional model to be verified;

[0055] Step S43: Perform virtual performance index evaluation on the three-dimensional model to be verified to obtain a simulation performance report;

[0056] Step S44: Perform model iterative optimization decision on the simulation performance report to obtain an adaptive three-dimensional model.

[0057] The present invention ensures that the generated three-dimensional model meets the preset design constraint conditions by applying the optimization design constraint to the pre-built initial parameterized model and performing constraint-driven parameter assignment, and converts the design constraint into an operable parameterized model, laying a foundation for the subsequent adaptive generation of geometric forms. The three-dimensional model to be verified is automatically generated according to the constraint assignment model, which realizes the automation of the design process, improves the design efficiency, and can quickly generate a variety of design schemes that meet the design constraints, providing more options for subsequent performance evaluation and optimization. By evaluating the virtual performance index of the generated three-dimensional model to be verified, such as using finite element analysis to evaluate its mechanical properties, thermal properties, etc., the performance of the model under actual working conditions can be predicted in advance, and potential design defects can be identified, providing guidance for subsequent model optimization, saving time and cost. By analyzing the simulation performance report and making iterative optimization decisions based on the evaluation results, the design scheme can be continuously improved until all performance indicators meet the design requirements, and finally a performance-optimized adaptive three-dimensional model is obtained, which can better meet the needs of actual working conditions and effectively reduce the risk of failure. The iterative optimization process ensures the performance and reliability of the final model.

[0058] Preferably, step S5 comprises the following steps:

[0059] Step S51: monitoring the service status of the new model on the adaptive three-dimensional model to obtain the operation history of the new model;

[0060] Step S52: Compare the actual performance data of the new model operation history with the behavior feature vector to obtain a performance difference report;

[0061] Step S53: recording maintenance events and fault information of the adaptive three-dimensional model to obtain a new model maintenance file;

[0062] Step S54: Calculate and evaluate the service performance index according to the performance difference report and the new model maintenance file to obtain a preliminary performance index set;

[0063] Step S55: Perform design optimization strategy feedback on the preliminary performance indicator set to obtain service performance indicators.

[0064] The present invention monitors the service status of the adaptive three-dimensional model parts put into actual use and records their operation data, constructs the operation history of the new model, provides actual operation data for subsequent performance evaluation and optimization, and makes up for the gap between simulation analysis and actual working conditions. By comparing and analyzing the operation data of the new model with the historical data (behavior feature vector), the performance improvement effect of the new model can be quantitatively evaluated, such as whether the vibration level is reduced, whether the temperature change is more stable, etc., to provide data support for design optimization and verify the effectiveness of the previous steps. By recording the maintenance events and fault information of the new model, a new model maintenance archive is constructed, which provides data support for evaluating the reliability and maintenance cost of the new model, and can more accurately calculate key indicators such as MTBF and MTTR. By combining the data in the performance difference report and the new model maintenance archive, various service performance indicators of the new model, such as MTBF, MTTR, maintenance cost, etc., are calculated, forming a preliminary performance indicator set, which provides a quantitative basis for the final performance evaluation and design optimization strategy feedback. By analyzing a preliminary set of performance indicators and feeding the evaluation results back to the design optimization stage, a closed-loop feedback optimization mechanism is formed, which can continuously improve the design methods and models, thereby continuously improving the performance and reliability of components, reducing maintenance costs, and enabling the design scheme to better adapt to actual working conditions.

[0065] Preferably, the present invention further provides an intelligent three-dimensional model design system for spare parts, which is used to execute the intelligent three-dimensional model design method for spare parts as described above, and the intelligent three-dimensional model design system for spare parts includes:

[0066] The dynamic feature extraction module is used to obtain multi-dimensional perception data of spare parts to obtain the original sensor data stream; to mine the associated feature parameters of the original sensor data stream to obtain the primary feature parameter set; to perform multi-dimensional feature fusion construction on the primary feature parameter set to obtain the behavior feature vector;

[0067] The risk spectrum construction module is used to obtain the historical failure data of spare parts; train the multimodal failure prediction model based on the historical failure data of spare parts to obtain the multimodal failure prediction model; use the multimodal failure prediction model to deduce the future failure probability and accumulate the time series failure risk to obtain the failure probability distribution;

[0068] The topological configuration evolution module is used to analyze the association between failure mechanism and design parameters of failure probability distribution to obtain a failure-parameter association table; to evaluate the sensitivity of design parameters on the failure-parameter association table to obtain parameter sensitivity evaluation data; to construct and optimize multi-objective constraints based on the parameter sensitivity evaluation data to obtain optimized design constraints;

[0069] The parameterized kernel reconstruction module is used to perform constraint-driven parameter assignment according to the optimization design constraint to obtain a constraint assignment model; and to construct a geometric morphology adaptive three-dimensional model according to the constraint assignment model to obtain an adaptive three-dimensional model;

[0070] The closed-loop performance evaluation module is used to monitor the service status of the new model of the adaptive three-dimensional model, generate a maintenance file, and obtain a new model maintenance file; and generate a service performance indicator according to the new model maintenance file to obtain a service performance indicator. BRIEF DESCRIPTION OF THE DRAWINGS

[0071] Figure 1 A schematic diagram of a process flow of an intelligent three-dimensional model design method for spare parts;

[0072] Figure 2 Detailed implementation flow chart of step S1 in the present invention;

[0073] Figure 3 It is a schematic diagram of the detailed implementation steps of step S3 in the present invention.

[0074] The purpose, features and advantages of the present invention will be further described with reference to the accompanying drawings in conjunction with the embodiments. DETAILED DESCRIPTION

[0075] The technical method of the present invention is described clearly and completely below in conjunction with the accompanying drawings. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by technicians in this field without creative work are within the scope of protection of the present invention.

[0076] In addition, the accompanying drawings are only schematic illustrations of the present invention and are not necessarily drawn to scale. The same reference numerals in the figures represent the same or similar parts, and their repeated description will be omitted. Some of the block diagrams shown in the accompanying drawings are functional entities and do not necessarily correspond to physically or logically independent entities. The functional entities can be implemented in software form, or implemented in one or more hardware modules or integrated circuits, or implemented in different networks and / or processor methods and / or microcontroller methods.

[0077] It should be understood that, although the terms "first", "second", etc. may be used herein to describe various units, these units should not be limited by these terms. These terms are used only to distinguish one unit from another unit. For example, without departing from the scope of the exemplary embodiments, the first unit may be referred to as the second unit, and similarly the second unit may be referred to as the first unit. The term "and / or" used herein includes any and all combinations of one or more of the listed associated items.

[0078] To achieve this, please refer to Figures 1 to 3 , an intelligent three-dimensional model design method for spare parts, comprising the following steps:

[0079] Step S1: acquiring multi-dimensional sensing data of spare parts to obtain original sensor data stream; mining associated feature parameters of the original sensor data stream to obtain a primary feature parameter set; and constructing multi-dimensional feature fusion on the primary feature parameter set to obtain a behavior feature vector;

[0080] Step S2: Obtain historical failure data of spare parts; train a multimodal failure prediction model based on the historical failure data of spare parts to obtain a multimodal failure prediction model; use the multimodal failure prediction model to deduce future failure probabilities and perform sequential failure risk accumulation to obtain a failure probability distribution;

[0081] Step S3: performing failure mechanism and design parameter correlation analysis on the failure probability distribution to obtain a failure-parameter correlation table; performing design parameter sensitivity evaluation on the failure-parameter correlation table to obtain parameter sensitivity evaluation data; performing multi-objective constraint construction and optimization based on the parameter sensitivity evaluation data to obtain optimized design constraints;

[0082] Step S4: assigning constraint-driven parameters according to the optimization design constraints to obtain a constraint assignment model; constructing a geometric morphology adaptive three-dimensional model according to the constraint assignment model to obtain an adaptive three-dimensional model;

[0083] Step S5: monitor the new model service status of the adaptive three-dimensional model, and generate a maintenance file to obtain a new model maintenance file; generate a service performance index according to the new model maintenance file to obtain a service performance index.

[0084] In the embodiment of the present invention, reference Figure 1 FIG. 1 is a schematic diagram of the steps of the intelligent three-dimensional model design method for spare parts of the present invention. In this example, the intelligent three-dimensional model design method for spare parts includes the following steps:

[0085] Step S1: acquiring multi-dimensional sensing data of spare parts to obtain original sensor data stream; mining associated feature parameters of the original sensor data stream to obtain a primary feature parameter set; and constructing multi-dimensional feature fusion on the primary feature parameter set to obtain a behavior feature vector;

[0086] In an embodiment of the present invention, raw data streams are acquired from multi-source sensors at key locations of equipment. After preprocessing such as synchronous calibration and noise suppression, characteristic parameters related to the operating status of spare parts are extracted, such as vibration spectrum characteristics, temperature change rate, current statistical characteristics, etc. These characteristics are finally fused into a multi-dimensional behavioral feature vector to provide a data basis for subsequent failure prediction.

[0087] Step S2: Obtain historical failure data of spare parts; train a multimodal failure prediction model based on the historical failure data of spare parts to obtain a multimodal failure prediction model; use the multimodal failure prediction model to deduce future failure probabilities and perform sequential failure risk accumulation to obtain a failure probability distribution;

[0088] In the embodiment of the present invention, the historical fault data of spare parts is first obtained, and then cleaned and standardized to form a standardized fault file. Then, the behavior feature vector is temporally associated and annotated with the fault file to construct a fault annotation feature set. The data set is used to train a multimodal failure prediction model based on GBDT to predict the probability of different failure modes of spare parts in the future, and the final failure probability distribution is obtained by accumulating the time series failure risk.

[0089] Step S3: performing failure mechanism and design parameter correlation analysis on the failure probability distribution to obtain a failure-parameter correlation table; performing design parameter sensitivity evaluation on the failure-parameter correlation table to obtain parameter sensitivity evaluation data; performing multi-objective constraint construction and optimization based on the parameter sensitivity evaluation data to obtain optimized design constraints;

[0090] In the embodiment of the present invention, the failure probability distribution is firstly subjected to high-risk failure screening to screen out failure modes with a high probability of occurrence. Then, based on the engineering knowledge base and simulation analysis, the correlation between the failure mechanism and the design parameters is established, and the sensitivity of the design parameters to the failure probability is evaluated. Finally, the initial design constraints are generated according to the sensitivity analysis results, and the constraints are optimized and integrated through a multi-objective optimization algorithm to obtain the final optimized design constraints, which are used to guide the subsequent generation of the three-dimensional model.

[0091] Step S4: assigning constraint-driven parameters according to the optimization design constraints to obtain a constraint assignment model; constructing a geometric morphology adaptive three-dimensional model according to the constraint assignment model to obtain an adaptive three-dimensional model;

[0092] In the embodiment of the present invention, the optimization design constraint is applied to the pre-built parametric model, and the constraint-driven parameter assignment is performed to obtain the constraint assignment model. Then, the three-dimensional model to be verified is automatically generated according to the constraint assignment model. The virtual performance index of the model is evaluated by simulation methods such as finite element analysis, and a simulation performance report is generated. Finally, the model is iteratively optimized according to the simulation results until all performance indicators meet the design requirements, and the final adaptive three-dimensional model is obtained.

[0093] Step S5: monitoring the service status of the new model of the adaptive three-dimensional model, and generating a maintenance file to obtain a maintenance file of the new model; generating a service performance index according to the maintenance file of the new model to obtain a service performance index;

[0094] In the embodiment of the present invention, the actual service status of the spare parts manufactured based on the adaptive three-dimensional model is monitored, the operation data is collected, and the maintenance events and fault information are recorded to form the operation history and maintenance file of the new model. The operation data of the new model is compared and analyzed with the historical data to evaluate the performance difference. The service performance indicators such as MTBF, MTTR, maintenance cost, etc. are calculated based on the performance difference report and maintenance file. Finally, the evaluation results are fed back to the design optimization stage to form a closed-loop feedback optimization mechanism to guide the subsequent model improvement and design optimization.

[0095] Preferably, step S1 comprises the following steps:

[0096] Step S11: acquiring multi-dimensional sensing data of spare parts at key parts of the equipment to obtain original sensor data stream;

[0097] Step S12: performing heterogeneous data synchronization calibration on the original sensor data stream to obtain a synchronous calibration data set;

[0098] Step S13: performing noise interference identification and suppression on the synchronous calibration data set to obtain high signal-to-noise ratio data;

[0099] Step S14: mining associated feature parameters for high signal-to-noise ratio data to obtain a primary feature parameter set;

[0100] Step S15: construct a multi-dimensional feature fusion on the primary feature parameter set to obtain a behavior feature vector.

[0101] As an example of the present invention, refer to Figure 2 As shown, in this example, step S1 includes:

[0102] Step S11: acquiring multi-dimensional sensing data of spare parts at key parts of the equipment to obtain original sensor data stream;

[0103] In an embodiment of the present invention, various types of sensors are installed on key parts of the equipment, such as bearing seats, gear boxes, motor housings, etc., which are prone to failure or need to monitor the status. These sensors include but are not limited to: Pt100 thermal resistors for measuring temperature, piezoelectric acceleration sensors for measuring vibration, photoelectric encoders for measuring rotation speed, Hall sensors for measuring current, etc. Each sensor is connected to a data collector through a data line, and the data collector synchronously collects data from each sensor according to a preset sampling frequency (e.g., 1kHz), and stores the data in a database after being marked with a timestamp and a sensor ID, forming an original sensor data stream. For example, on the bearing seat of a rotating machine, a Pt100 thermal resistor and a piezoelectric acceleration sensor are installed at the same time, which are used to measure the bearing temperature and vibration acceleration, respectively. The data collector synchronously collects the data of these two sensors at a frequency of 1kHz, and stores the data in the database in the form of a timestamp, a sensor ID (e.g., "temperature sensor 1", "vibration sensor 1") and a numerical value, forming an original sensor data stream.

[0104] Step S12: performing heterogeneous data synchronization calibration on the original sensor data stream to obtain a synchronous calibration data set;

[0105] In an embodiment of the present invention, the raw sensor data stream is read from a database. First, the data of different sensors are time synchronized according to the timestamp of each data point. The sensor data of different sampling frequencies are aligned by linear interpolation method to ensure that all data points correspond to the same timestamp. Then, according to the calibration parameters of each sensor, the original voltage or current signal is converted into the corresponding physical quantity. For example, the voltage signal of the Pt100 thermal resistor is converted into Celsius temperature, and the voltage signal of the piezoelectric acceleration sensor is converted into g value (multiple of gravity acceleration). Finally, the calibrated data is arranged in the order of timestamps to form a synchronous calibration data set. For example, the voltage signal of "temperature sensor 1" is converted into Celsius temperature according to its calibration curve, and the voltage signal of "vibration sensor 1" is converted into g value according to its sensitivity. The converted temperature and vibration data are stored together with the corresponding timestamps to form a synchronous calibration data set.

[0106] Step S13: performing noise interference identification and suppression on the synchronous calibration data set to obtain high signal-to-noise ratio data;

[0107] In an embodiment of the present invention, data is read from a synchronous calibration data set. For each type of sensor data, a denoising algorithm that matches its characteristics is used to suppress noise. For example, for temperature data, a sliding average filtering method is used to remove high-frequency noise; for vibration data, a wavelet threshold denoising method is used to remove random noise; for current data, a median filtering method is used to remove pulse noise. The parameters of the denoising algorithm, such as the sliding average window size, wavelet basis function, threshold, etc., are adjusted according to the noise characteristics of the specific data to suppress the noise to the greatest extent while retaining useful signal components. For example, a sliding average filter of 5 data points is used for the bearing temperature data, and a db4 wavelet basis function is used for 5-layer wavelet decomposition of the vibration acceleration data, and soft threshold denoising is performed on the detail coefficients. The processed data is stored as high signal-to-noise ratio data.

[0108] Step S14: mining associated feature parameters for high signal-to-noise ratio data to obtain a primary feature parameter set;

[0109] In an embodiment of the present invention, characteristic parameters related to the operating status and potential failure modes of spare parts are extracted from high signal-to-noise ratio data. Different feature extraction methods are used for different types of sensor data. For example, for vibration data, a fast Fourier transform (FFT) is performed to obtain the spectrum, and spectrum features such as peak frequency, root mean square value, frequency band energy, etc. are extracted; for temperature data, time domain features such as temperature change rate and temperature gradient are calculated; for current data, statistical features such as the average value, variance, and peak-to-peak value of the current are calculated. The various extracted characteristic parameters are stored as a primary characteristic parameter set. For example, an FFT is performed on the bearing vibration data to extract the peak frequency, root mean square value, and frequency band energy in the 10Hz-1kHz frequency band. The temperature change rate per minute is calculated for the bearing temperature data. These characteristic parameters are stored as a primary characteristic parameter set.

[0110] Step S15: constructing a multi-dimensional feature fusion on the primary feature parameter set to obtain a behavior feature vector;

[0111] In an embodiment of the present invention, all extracted feature parameters are read from the primary feature parameter set. Different types of feature parameters are standardized, for example, the value of each feature parameter is scaled to between 0 and 1 to eliminate the influence of different feature dimensions and numerical ranges. Then, all standardized feature parameters are combined into a multidimensional vector, namely a behavioral feature vector, in a predefined order. The dimension of the behavioral feature vector is equal to the number of feature parameters. For example, after the bearing vibration characteristics (peak frequency, root mean square value, frequency band energy) and temperature characteristics (temperature change rate) are standardized, they are combined in order into a five-dimensional vector as a behavioral feature vector that describes the operating state of the bearing.

[0112] Preferably, step S14 comprises the following steps:

[0113] Step S141: constructing a mechanical model according to the high signal-to-noise ratio data to obtain a mechanical model;

[0114] Step S142: performing stress history deduction based on high signal-to-noise ratio data and a mechanical model to obtain a stress time series;

[0115] Step S143: extracting the cyclic load spectrum of the stress time series to obtain a cyclic characteristic spectrum;

[0116] Step S144: Calculate fatigue damage parameters according to the cycle characteristic spectrum to obtain fatigue damage index;

[0117] Step S145: performing common feature integration on fatigue damage indicators to obtain a primary feature parameter set.

[0118] In an embodiment of the present invention, the finite element analysis software Abaqus is used to establish a mechanical model of the spare parts according to their specific structure and stress conditions. The model includes the geometric shape, material properties (such as elastic modulus, Poisson's ratio, density), mesh division and boundary conditions of the spare parts. For example, for a shaft subjected to bending load, its three-dimensional geometric model is created in Abaqus, the material property is defined as 45 steel, the mesh is divided using a hexahedral mesh, and fixed constraints are set at both ends of the shaft as boundary conditions. The input parameters of the model include a load time series obtained from high signal-to-noise ratio data, for example, strain data obtained from a strain gauge sensor installed on the shaft, and converted into load data. The mechanical model is used for subsequent stress history deduction.

[0119] The load time series extracted from the high signal-to-noise ratio data is used as input and applied to the finite element model established in step S141. Perform static or dynamic analysis in Abaqus to calculate the stress distribution of the model at each time step. Extract the stress values ​​of key parts of the spare parts (such as stress concentration areas) and arrange them in chronological order to form a stress time series. For example, the load time series obtained from the strain gauge sensor is applied to the finite element model of the shaft, and a transient dynamic analysis is performed in Abaqus to calculate the stress distribution of the shaft at each time step. Extract the maximum principal stress value at the chamfer of the shaft shoulder to form a stress time series.

[0120] The cyclic load spectrum of the stress time series obtained in step S142 is extracted by using the rain flow counting method. The stress peaks and valleys in the stress time series are arranged in chronological order to identify complete stress cycles. The amplitude and mean of each stress cycle are counted, and they are grouped according to the amplitude size. The number of cycles occurring in each amplitude range is counted to form a cyclic characteristic spectrum. The cyclic characteristic spectrum is presented in the form of a table or a histogram, describing the number of occurrences of different stress amplitudes and means. For example, the rain flow counting method is applied to the stress time series at the chamfer of the shoulder to extract the amplitude and mean of each stress cycle. The stress amplitudes are grouped at intervals of 10 MPa, and the number of cycles occurring in each amplitude range is counted to form a cyclic characteristic spectrum.

[0121] According to the SN curve (stress-life curve) of the material and the cyclic characteristic spectrum obtained in step S143, the fatigue damage parameters are calculated. For example, the Miner linear cumulative damage theory is used to calculate the damage value caused by each stress cycle, and the damage values ​​of all cycles are accumulated to obtain the total fatigue damage accumulation. In addition, the equivalent stress amplitude can also be calculated, and the variable amplitude load spectrum is equivalent to the constant amplitude load for fatigue life prediction. For example, according to the SN curve and cyclic characteristic spectrum of No. 45 steel, the damage value caused by each stress cycle is calculated, and all damage values ​​are accumulated to obtain the fatigue damage accumulation at the chamfer of the shoulder. At the same time, the equivalent stress amplitude is calculated for subsequent fatigue life prediction.

[0122] The fatigue damage index (e.g., fatigue damage accumulation, equivalent stress amplitude) calculated in step S144 is integrated with the common characteristic parameters (e.g., vibration frequency, temperature change rate, etc.) extracted from other sensor data in step S14 to form a final primary characteristic parameter set. Before integration, different types of characteristic parameters are standardized, such as MinMaxScaler or Z-score standardization, to eliminate the influence of dimension and numerical range. For example, characteristic parameters such as fatigue damage accumulation, equivalent stress amplitude, vibration peak frequency, temperature change rate, etc. are standardized and integrated together to form a final primary characteristic parameter set.

[0123] Preferably, step S2 comprises the following steps:

[0124] Step S21: Acquire historical fault data of spare parts; integrate historical fault information of spare parts historical fault data to obtain standardized fault files;

[0125] Step S22: labeling the behavioral feature vector and the standardized fault file in association with the behavioral feature to obtain a fault labeling feature set;

[0126] Step S23: training a multimodal failure prediction model using the fault annotation feature set to obtain a multimodal failure prediction model;

[0127] Step S24: inputting the behavior feature vector data into the multi-modal failure prediction model to deduce the future failure probability and obtain the immediate failure probability;

[0128] Step S25: Accumulate the instantaneous failure probability in terms of sequential failure risk to obtain a failure probability distribution.

[0129] In an embodiment of the present invention, historical fault data of target spare parts are collected from the enterprise's maintenance management system, fault record database, and maintenance personnel's reports. These data contain information such as the time of occurrence of the fault, fault description, failure mode classification (such as fatigue fracture, wear failure, corrosion failure, etc.), maintenance and replacement records, etc. The collected data is cleaned and preprocessed, such as removing duplicate records, filling missing values, unifying fault description terms, etc. The processed data is arranged in chronological order and stored as a structured standardized fault file, such as in the form of CSV or database tables. Each fault record contains a unique ID, occurrence time, failure mode classification, and related sensor data.

[0130] Each behavior feature vector generated in step S1 is temporally associated with the fault record in the standardized fault file. For each fault record, the behavior feature vector within a time window before the fault occurred (for example, 1 hour before the fault occurred) is extracted. These behavior feature vectors are marked with corresponding failure mode labels. For example, if the behavior feature vector within a certain time period corresponds to a fatigue fracture failure, the vector is marked as "fatigue fracture". If no failure occurs within a period of time, the corresponding behavior feature vector is marked as "normal". All behavior feature vectors marked with failure mode labels are collected together to form a fault annotation feature set.

[0131] Use the fault annotation feature set to train a multimodal failure prediction model based on a gradient boosted decision tree (GBDT). Divide the fault annotation feature set into a training set and a test set. Use the training set data to train the GBDT model. The input features of the model are behavioral feature vectors, and the output is the probability of different failure modes. Adjust the parameters of the GBDT model, such as the learning rate, tree depth, number of trees, etc., to optimize the prediction performance of the model, and use the test set to evaluate the accuracy and recall of the model. After the training is completed, save the trained GBDT model, that is, the multimodal failure prediction model.

[0132] The latest behavior feature vector generated in step S1 is input into the multimodal failure prediction model trained in step S23. The model will output a vector containing the predicted probabilities of different failure modes. For example, the model outputs [0.01, 0.05, 0.9], indicating that the probabilities of fatigue fracture, wear failure, and normal state of the current spare parts are 0.01, 0.05, and 0.9, respectively. This output vector is the instantaneous failure probability.

[0133] The immediate failure probability within a period of time (for example, the next 24 hours) is accumulated to obtain the failure probability distribution. For each failure mode, the immediate failure probability of all time points within the time period is averaged to obtain the average failure probability of the failure mode within the time period. More complex accumulation methods can also be used, such as considering the time decay effect of the failure probability, or setting different weights according to different failure modes. The final failure probability distribution describes the probability of different failure modes occurring in the future and their time distribution.

[0134] Preferably, step S3 comprises the following steps:

[0135] Step S31: performing high-risk failure screening on the failure probability distribution to obtain a significant failure mode set;

[0136] Step S32: performing correlation analysis between failure mechanism and design parameters on the significant failure mode set to obtain a failure-parameter correlation table;

[0137] Step S33: performing design parameter sensitivity evaluation on the failure-parameter association table to obtain parameter sensitivity evaluation data;

[0138] Step S34: generating initial design constraints according to the parameter sensitivity evaluation data to obtain an initial design constraint set;

[0139] Step S35: performing multi-objective constraint optimization and fusion on the initial design constraint set to obtain optimized design constraints.

[0140] As an example of the present invention, refer to Figure 3 As shown, in this example, step S3 includes:

[0141] Step S31: performing high-risk failure screening on the failure probability distribution to obtain a significant failure mode set;

[0142] In an embodiment of the present invention, the average failure probability of each failure mode in the future period of time (for example, the next 24 hours) is extracted from the failure probability distribution obtained in step S2. A threshold value is preset, such as 0.1. Failure modes whose average failure probability exceeds the threshold are added to the significant failure mode set. For example, if the average failure probability of fatigue fracture is 0.2, which exceeds the threshold value of 0.1, "fatigue fracture" is added to the significant failure mode set. If the average failure probability of wear failure is 0.05, which is lower than the threshold value of 0.1, it is not added to the significant failure mode set. The significant failure mode set contains failure modes that need to be focused on and optimized.

[0143] Step S32: performing correlation analysis between failure mechanism and design parameters on the significant failure mode set to obtain a failure-parameter correlation table;

[0144] In the embodiment of the present invention, for each failure mode in the significant failure mode set, the correlation between the failure mechanism and the design parameters is established based on the knowledge of related disciplines such as material mechanics, fatigue theory, and tribology. For example, for fatigue fracture, its failure mechanism is related to design parameters such as stress concentration, material fatigue strength, and surface roughness; for wear failure, its failure mechanism is related to design parameters such as contact pressure, sliding speed, material hardness, and surface lubrication state. Each failure mode and its related design parameters are recorded in a failure-parameter association table in a tabular form. For example, in the failure-parameter association table, "fatigue fracture" is associated with parameters such as "stress concentration factor", "material fatigue limit", and "surface roughness Ra"; "wear failure" is associated with parameters such as "contact pressure", "sliding speed", "material hardness HV", and "lubricating oil viscosity".

[0145] Step S33: performing design parameter sensitivity evaluation on the failure-parameter association table to obtain parameter sensitivity evaluation data;

[0146] In an embodiment of the present invention, for each design parameter in the failure-parameter association table, the degree of its influence on the probability of occurrence of the corresponding failure mode is evaluated. The finite element analysis method is used to observe the change in the failure probability by changing the value of the design parameter, such as increasing or decreasing by 5%. The ratio of the change in the failure probability to the change in the design parameter is used as the sensitivity of the design parameter. Each design parameter and its corresponding sensitivity value are recorded in the parameter sensitivity evaluation data. For example, the shoulder chamfer radius is increased by 5%, and the change in the failure probability of fatigue fracture is calculated using finite element analysis. The ratio of the change in the failure probability to the change in the chamfer radius is used as the sensitivity of the chamfer radius to fatigue fracture, and the result is recorded in the parameter sensitivity evaluation data.

[0147] Step S34: generating initial design constraints according to the parameter sensitivity evaluation data to obtain an initial design constraint set;

[0148] In an embodiment of the present invention, initial design constraints are formulated for each high-risk failure mode based on parameter sensitivity evaluation data. Design parameters that are more sensitive to failure modes are selected, and the constraint range is set according to the target value of the failure probability. For example, if the shoulder chamfer radius is highly sensitive to fatigue fracture, and the goal is to reduce the probability of fatigue fracture to below 0.01, a minimum value constraint for the chamfer radius can be set. Add all initial design constraints for high-risk failure modes to the initial design constraint set. For example, add "chamfer radius >= 5mm" to the initial design constraint set to reduce the risk of fatigue fracture.

[0149] Step S35: performing multi-objective constraint optimization and fusion on the initial design constraint set to obtain optimized design constraints;

[0150] In an embodiment of the present invention, a multi-objective optimization algorithm, such as the NSGA-II algorithm, is used to optimize the initial design constraint set. The optimization goal is to simultaneously reduce the probability of various failure modes and meet other design requirements, such as weight restrictions, cost restrictions, etc. The algorithm searches for a combination of design parameters that meets all constraints and generates a set of Pareto optimal solutions. Select an optimal solution from the Pareto optimal solution and convert it into the final optimization design constraint. For example, the initial design constraint set is optimized using the NSGA-II algorithm, with the goal of minimizing the probability of fatigue fracture and wear failure and satisfying the constraint that the weight is less than 1kg. Select an optimal solution from the generated Pareto optimal solution and convert it into the final optimization design constraint, such as "chamfer radius>=5mm", "material hardness HV>=400", and "surface roughness Ra<=0.8μm".

[0151] Preferably, step S32 includes the following steps:

[0152] Step S321: Perform preliminary knowledge-based parameter identification on the significant failure mode set to obtain a preliminary associated parameter list;

[0153] Step S322: quantifying the parameter impact driven by simulation on the preliminary associated parameter list to obtain parameter sensitivity data;

[0154] Step S323: experimentally verifying the parameter sensitivity data and calibrating the correlation relationship to obtain calibrated parameter sensitivity;

[0155] Step S324: Develop a failure mechanism association rule set based on the preliminary association parameter list and the calibrated parameter sensitivity.

[0156] In an embodiment of the present invention, for each failure mode in the significant failure mode set, the knowledge base and literature in related fields such as material mechanics, fatigue theory, tribology, etc. are queried to identify the preliminary design parameters related to the failure mode. For example, for the significant failure mode "fatigue fracture", by querying the material mechanics knowledge base, the parameters related to fatigue life, such as stress concentration factor, material fatigue limit, surface roughness, load size, etc. are identified. These parameters are recorded in the preliminary associated parameter list to form a parameter list associated with "fatigue fracture". Similarly, for other significant failure modes, such as "wear failure", "corrosion failure", etc., the same knowledge base query and parameter identification are also performed to generate corresponding preliminary associated parameter lists.

[0157] Finite element analysis software Abaqus is used to simulate and analyze each parameter in the preliminary associated parameter list to quantify its impact on the failure mode. For each parameter, a series of parameter values ​​are set, such as increasing or decreasing by 5%, 10%, 15%, etc. based on the baseline value. For each parameter value, simulation calculations are performed in Abaqus, such as fatigue analysis or wear analysis, to obtain corresponding failure indicators, such as fatigue life or wear rate. The different values ​​of each parameter and its corresponding failure indicator are recorded in the parameter sensitivity data, such as in the form of a table or chart, so as to facilitate subsequent analysis of the degree of influence of the parameter on the failure mode.

[0158] Experimental verification is performed on the parameter sensitivity data obtained in step S322. A series of test specimens are produced according to the parameter value range used in the simulation analysis. The same load and boundary conditions as in the simulation analysis are applied to each test specimen, and fatigue tests or wear tests are performed to obtain actual failure indicators, such as fatigue life or wear rate. The failure indicators obtained from the experiment are compared with the failure indicators obtained from the simulation analysis, and the deviation between the two is calculated. The parameter sensitivity data obtained in step S322 is calibrated according to the deviation value. For example, if the experimental results show that the sensitivity of a certain parameter is higher than the simulation results, the sensitivity value of the parameter is revised upward. The calibrated parameter sensitivity data more accurately reflects the degree of influence of the design parameters on the failure mode.

[0159] According to the preliminary associated parameter list obtained in step S321 and the calibrated parameter sensitivity obtained in step S323, a failure mechanism association rule set is established. The rule set represents the association relationship between parameters and failure modes in the form of "IF-THEN". For example, for fatigue fracture, if the calibrated parameter sensitivity data shows that the stress concentration factor has the greatest impact on fatigue life, the following rule can be established: "IF stress concentration factor increases THEN fatigue life decreases". Similarly, association rules between other parameters and fatigue life can be established, such as "IF surface roughness increases THEN fatigue life decreases". All rules related to failure modes are added to the failure mechanism association rule set to form a complete knowledge base to guide subsequent design optimization.

[0160] Preferably, step S33 includes the following steps:

[0161] Step S331: performing local sensitivity calculation on the failure-parameter association table to obtain local sensitivity data;

[0162] Step S332: performing a global sensitivity analysis according to the failure-parameter association table and the failure probability distribution to obtain a global sensitivity index;

[0163] Step S333: performing simulation-assisted sensitivity verification according to the failure-parameter association table to obtain simulation verification sensitivity data;

[0164] Step S334: Generate parameter sensitivity evaluation data for the local sensitivity data, the global sensitivity index, and the simulation verification sensitivity data to obtain parameter sensitivity evaluation data.

[0165] In an embodiment of the present invention, for each design parameter in the failure-parameter association table, a small disturbance is made near its baseline value, for example, ±5%. The partial derivative of the failure probability with respect to the parameter is calculated using the finite difference method. Specifically, the parameter value is increased by 5% and decreased by 5%, and the corresponding failure probabilities are calculated respectively. The difference in the failure probabilities obtained by the two calculations is divided by the change in the parameter value (10%) to obtain the local sensitivity of the parameter at the baseline value. The local sensitivity of each design parameter is recorded in the local sensitivity data, for example, in a table form, where each row represents a design parameter and each column represents a failure mode.

[0166] Based on the failure-parameter association table and failure probability distribution, the Sobol method is used to perform global sensitivity analysis. The Sobol method evaluates the global sensitivity of the parameters by analyzing the variance contribution to the failure probability when the parameters change within the value range. First, a set of parameter samples is generated according to the parameter range defined in the failure-parameter association table. Then, the failure probability corresponding to each parameter sample is calculated using the multimodal failure prediction model trained in step S2. Finally, the first-order Sobol index and total Sobol index of each parameter are calculated based on these failure probabilities, representing the direct impact and overall impact of the parameter, respectively. The calculated Sobol index is used as a global sensitivity index and stored in the global sensitivity index data.

[0167] For the design parameters and failure modes listed in the failure-parameter association table, the finite element analysis software Abaqus is used to perform simulation analysis to verify the calculation results of parameter sensitivity. For each design parameter, a perturbation is performed near its baseline value, such as ±10%. A finite element model of the spare parts is established in Abaqus, and the model parameters are modified according to the perturbation value of the parameter. Simulation calculations, such as fatigue analysis or wear analysis, are performed to obtain failure indicators under different parameter values, such as fatigue life or wear rate. The sensitivity of the parameters is calculated based on the simulation results, and the results are recorded in the simulation verification sensitivity data.

[0168] The local sensitivity data obtained in step S331, the global sensitivity index obtained in step S332, and the simulation verification sensitivity data obtained in step S333 are comprehensively considered to generate the final parameter sensitivity evaluation data. For example, the three types of sensitivity data can be fused by weighted averaging. The weights can be set according to the reliability and importance of the data. For example, if the reliability of the experimental verification data is the highest, a higher weight is assigned to it. The fused sensitivity data is stored in the parameter sensitivity evaluation data for subsequent initial design constraint generation. The parameter sensitivity evaluation data contains the comprehensive sensitivity value of each design parameter to each failure mode.

[0169] Preferably, step S4 comprises the following steps:

[0170] Step S41: applying the optimization design constraint to the pre-built initial parameterized model, performing constraint-driven parameter assignment, and obtaining a constraint assignment model;

[0171] Step S42: performing geometric shape adaptive generation according to the constraint assignment model to obtain a three-dimensional model to be verified;

[0172] Step S43: Perform virtual performance index evaluation on the three-dimensional model to be verified to obtain a simulation performance report;

[0173] Step S44: Perform model iterative optimization decision on the simulation performance report to obtain an adaptive three-dimensional model.

[0174] In an embodiment of the present invention, an initial parametric model of a spare part is pre-constructed, the model is created using the CAD software SolidWorks, and includes adjustable parameters related to the optimization design constraints, such as chamfer radius, hole diameter, material thickness, etc. The parameter values ​​in the optimization design constraints obtained in step S35 are assigned to the corresponding parameters in the initial parametric model. For example, if the optimization design constraints stipulate that the chamfer radius must be greater than or equal to 5 mm, the chamfer radius parameter in the parametric model is set to 5 mm or greater. After completing the parameter assignment, the parametric model is updated to obtain a constraint assignment model. The geometric shape and material properties of the model have been adjusted according to the optimization design constraints.

[0175] The automatic reconstruction function of the parametric modeling software SolidWorks is used to generate a new three-dimensional geometric model based on the constraint assignment model obtained in step S41. Specifically, SolidWorks automatically adjusts the shape and size of the model based on the parameter values ​​and geometric relationships defined in the constraint assignment model to generate a three-dimensional model that meets the optimization design constraints. The generated model is saved as a three-dimensional model to be verified, which will be used for subsequent performance evaluation and iterative optimization.

[0176] Import the three-dimensional model to be verified generated in step S42 into the finite element analysis software Abaqus. Set the load and boundary conditions for the simulation analysis according to the actual working conditions of the spare parts. For example, for an axis subjected to bending load, fixed constraints are applied at both ends of the axis, and bending force is applied in the middle of the axis. Perform finite element analysis to calculate the stress, strain and displacement distribution of the model under load. Extract performance indicators such as stress values ​​and maximum deformation of key parts, and compare these indicators with predefined design requirements, for example, whether the maximum stress value is less than the yield strength of the material, whether the maximum deformation meets the accuracy requirements, etc. The simulation analysis results are organized into a simulation performance report, which contains the specific values ​​of each performance indicator and the evaluation results of whether the design requirements are met.

[0177] Analyze the simulation performance report obtained in step S43. If all performance indicators meet the design requirements, the current three-dimensional model to be verified is determined as the final adaptive three-dimensional model. If there are performance indicators that do not meet the design requirements, it is necessary to adjust the optimization design constraints in step S35 according to the specific results in the simulation performance report. For example, if the simulation results show that the maximum stress value exceeds the yield strength of the material, it is necessary to increase the thickness of the key parts of the spare parts or change the material properties. Reapply the adjusted optimization design constraints to step S41, re-assign parameters, generate geometry, and evaluate performance until all performance indicators meet the design requirements. The final three-dimensional model that meets all design requirements is the adaptive three-dimensional model.

[0178] Preferably, step S5 comprises the following steps:

[0179] Step S51: monitoring the service status of the new model on the adaptive three-dimensional model to obtain the operation history of the new model;

[0180] Step S52: Compare the actual performance data of the new model operation history with the behavior feature vector to obtain a performance difference report;

[0181] Step S53: recording maintenance events and fault information of the adaptive three-dimensional model to obtain a new model maintenance file;

[0182] Step S54: Calculate and evaluate the service performance index according to the performance difference report and the new model maintenance file to obtain a preliminary performance index set;

[0183] Step S55: Perform design optimization strategy feedback on the preliminary performance indicator set to obtain service performance indicators.

[0184] In an embodiment of the present invention, spare parts manufactured based on the adaptive three-dimensional model are installed in actual equipment. During the operation of the equipment, the same sensors and data acquisition system as step S1 are used to collect the operating data of the spare parts in real time, such as temperature, vibration, strain, current, etc. At the same time, the operating time, workload, ambient temperature and other information of the spare parts are recorded. The collected data is stored in chronological order, and the unique identifier and installation time of the spare parts are added to form a new model operation history. The new model operation history contains a complete operation data record of the spare parts in actual service status.

[0185] Extract the same characteristic parameters as those in step S1 from the new model operation history, such as vibration spectrum characteristics, temperature change rate, etc., to form a new behavior feature vector. Compare and analyze the new behavior feature vector with the behavior feature vector constructed based on historical data in step S1. Use statistical analysis methods, such as t-test or Mann-Whitney U test, to compare the differences between the two groups of behavior feature vectors in various dimensions and calculate the significance level of the differences. Record the comparative analysis results in the performance difference report, including the difference values, significance levels, and corresponding statistical charts of each characteristic parameter.

[0186] During the entire service life of a spare part, all maintenance events and fault information related to the spare part are recorded. Maintenance events include regular inspection, lubrication, cleaning, etc. Fault information includes fault occurrence time, fault type, fault cause, repair measures, etc. This information is recorded in the new model maintenance file and associated with the unique identifier of the spare part. The new model maintenance file is used to evaluate the reliability and maintenance cost of spare parts.

[0187] Based on the performance difference report and the new model maintenance file, calculate the various service performance indicators of spare parts, such as mean time between failures (MTBF), mean time to repair (MTTR), maintenance cost, service life, etc. For example, calculate MTBF based on the number of failures and operating time recorded in the new model maintenance file; calculate MTTR based on the repair time record; calculate maintenance cost based on the cost of replacing parts; calculate service life based on the retirement time of spare parts. Summarize the calculated performance indicators into a preliminary performance indicator set.

[0188] Analyze each indicator in the preliminary performance indicator set and compare it with the pre-set target value. For example, compare whether the MTBF of the new model is higher than the MTBF of the old model, whether the maintenance cost is reduced, etc. If all indicators reach or exceed the target value, the current design scheme is considered to be effective. If some indicators do not reach the target value, it is necessary to analyze the reasons for non-compliance based on the performance difference report and the new model maintenance file, and put forward improvement suggestions. For example, if the MTBF is lower than expected, it is necessary to analyze the main cause of the failure and adjust the design parameters or maintenance strategies in a targeted manner. Feedback the analysis results and improvement suggestions to step S3 for subsequent design optimization to form a closed-loop feedback mechanism. The final output service performance indicators include the final evaluation results of each performance indicator and targeted optimization strategy recommendations.

[0189] Preferably, the present invention further provides an intelligent three-dimensional model design system for spare parts, which is used to execute the intelligent three-dimensional model design method for spare parts as described above, and the intelligent three-dimensional model design system for spare parts includes:

[0190] The dynamic feature extraction module is used to obtain multi-dimensional perception data of spare parts to obtain the original sensor data stream; to mine the associated feature parameters of the original sensor data stream to obtain the primary feature parameter set; to perform multi-dimensional feature fusion construction on the primary feature parameter set to obtain the behavior feature vector;

[0191] The risk spectrum construction module is used to obtain the historical failure data of spare parts; train the multimodal failure prediction model based on the historical failure data of spare parts to obtain the multimodal failure prediction model; use the multimodal failure prediction model to deduce the future failure probability and accumulate the time series failure risk to obtain the failure probability distribution;

[0192] The topological configuration evolution module is used to analyze the association between failure mechanism and design parameters of failure probability distribution to obtain a failure-parameter association table; to evaluate the sensitivity of design parameters on the failure-parameter association table to obtain parameter sensitivity evaluation data; to construct and optimize multi-objective constraints based on the parameter sensitivity evaluation data to obtain optimized design constraints;

[0193] The parameterized kernel reconstruction module is used to perform constraint-driven parameter assignment according to the optimization design constraint to obtain a constraint assignment model; and to construct a geometric morphology adaptive three-dimensional model according to the constraint assignment model to obtain an adaptive three-dimensional model;

[0194] The closed-loop performance evaluation module is used to monitor the service status of the new model of the adaptive three-dimensional model, generate a maintenance file, and obtain a new model maintenance file; and generate a service performance indicator according to the new model maintenance file to obtain a service performance indicator.

[0195] Therefore, the embodiments should be regarded as illustrative and non-restrictive from all points, and the scope of the present invention is limited by the appended claims rather than the above description, and it is therefore intended that all changes falling within the meaning and range of equivalent elements of the application documents are included in the present invention.

[0196] The above description is only a specific embodiment of the present invention, so that those skilled in the art can understand or implement the present invention. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but should conform to the widest scope consistent with the principles and novel features invented herein.

Claims

1. An intelligent three-dimensional model design method for spare parts, characterized in that: The following steps are involved: Step S1: acquiring multi-dimensional sensing data of spare parts to obtain original sensor data stream; mining associated feature parameters of the original sensor data stream to obtain a primary feature parameter set; and constructing multi-dimensional feature fusion on the primary feature parameter set to obtain a behavior feature vector; Step S2: Obtain historical failure data of spare parts; train a multimodal failure prediction model based on the historical failure data of spare parts to obtain a multimodal failure prediction model; use the multimodal failure prediction model to deduce future failure probabilities and perform sequential failure risk accumulation to obtain a failure probability distribution; Step S3: analyzing the association between failure mechanism and design parameters of the failure probability distribution to obtain a failure-parameter association table; Perform design parameter sensitivity evaluation on the failure-parameter association table to obtain parameter sensitivity evaluation data; Based on the parameter sensitivity evaluation data, multi-objective constraints are constructed and optimized to obtain the optimal design constraints; Step S4: assigning constraint-driven parameters according to the optimization design constraints to obtain a constraint assignment model; constructing a geometric morphology adaptive three-dimensional model according to the constraint assignment model to obtain an adaptive three-dimensional model; Step S5: monitor the new model service status of the adaptive three-dimensional model, and generate a maintenance file to obtain a new model maintenance file; generate a service performance index according to the new model maintenance file to obtain a service performance index.

2. The intelligent three-dimensional model design method for spare parts according to claim 1, characterized in that: Step S1 includes the following steps: Step S11: acquiring multi-dimensional sensing data of spare parts at key parts of the equipment to obtain original sensor data stream; Step S12: performing heterogeneous data synchronization calibration on the original sensor data stream to obtain a synchronous calibration data set; Step S13: performing noise interference identification and suppression on the synchronous calibration data set to obtain high signal-to-noise ratio data; Step S14: mining associated feature parameters for high signal-to-noise ratio data to obtain a primary feature parameter set; Step S15: construct a multi-dimensional feature fusion on the primary feature parameter set to obtain a behavior feature vector.

3. The intelligent three-dimensional model design method for spare parts according to claim 2, characterized in that: Step S14 includes the following steps: Step S141: constructing a mechanical model according to the high signal-to-noise ratio data to obtain a mechanical model; Step S142: performing stress history deduction based on high signal-to-noise ratio data and a mechanical model to obtain a stress time series; Step S143: extracting the cyclic load spectrum of the stress time series to obtain a cyclic characteristic spectrum; Step S144: Calculate fatigue damage parameters according to the cycle characteristic spectrum to obtain fatigue damage index; Step S145: performing common feature integration on fatigue damage indicators to obtain a primary feature parameter set.

4. The intelligent three-dimensional model design method for spare parts according to claim 1, characterized in that: Step S2 includes the following steps: Step S21: Acquire historical fault data of spare parts; integrate historical fault information of spare parts historical fault data to obtain standardized fault files; Step S22: labeling the behavioral feature vector and the standardized fault file in association with the behavioral feature to obtain a fault labeling feature set; Step S23: training a multimodal failure prediction model using the fault annotation feature set to obtain a multimodal failure prediction model; Step S24: inputting the behavior feature vector data into the multi-modal failure prediction model to deduce the future failure probability and obtain the immediate failure probability; Step S25: Accumulate the instantaneous failure probability in terms of sequential failure risk to obtain a failure probability distribution.

5. The intelligent three-dimensional model design method for spare parts according to claim 1, characterized in that: Step S3 includes the following steps: Step S31: performing high-risk failure screening on the failure probability distribution to obtain a significant failure mode set; Step S32: performing correlation analysis between failure mechanism and design parameters on the significant failure mode set to obtain a failure-parameter correlation table; Step S33: performing design parameter sensitivity evaluation on the failure-parameter association table to obtain parameter sensitivity evaluation data; Step S34: generating initial design constraints according to the parameter sensitivity evaluation data to obtain an initial design constraint set; Step S35: performing multi-objective constraint optimization and fusion on the initial design constraint set to obtain optimized design constraints.

6. The intelligent three-dimensional model design method for spare parts according to claim 5, characterized in that: Step S32 includes the following steps: Step S321: Perform preliminary knowledge-based parameter identification on the significant failure mode set to obtain a preliminary associated parameter list; Step S322: quantifying the parameter impact driven by simulation on the preliminary associated parameter list to obtain parameter sensitivity data; Step S323: experimentally verifying the parameter sensitivity data and calibrating the correlation relationship to obtain calibrated parameter sensitivity; Step S324: Develop a failure mechanism association rule set based on the preliminary association parameter list and the calibrated parameter sensitivity.

7. The intelligent three-dimensional model design method for spare parts according to claim 5, characterized in that: Step S33 includes the following steps: Step S331: performing local sensitivity calculation on the failure-parameter association table to obtain local sensitivity data; Step S332: performing a global sensitivity analysis according to the failure-parameter association table and the failure probability distribution to obtain a global sensitivity index; Step S333: performing simulation-assisted sensitivity verification according to the failure-parameter association table to obtain simulation-verified sensitivity data; Step S334: Generate parameter sensitivity evaluation data for the local sensitivity data, the global sensitivity index, and the simulation verification sensitivity data to obtain parameter sensitivity evaluation data.

8. The intelligent three-dimensional model design method for spare parts according to claim 1, characterized in that: Step S4 includes the following steps: Step S41: applying the optimization design constraint to the pre-built initial parameterized model, performing constraint-driven parameter assignment, and obtaining a constraint assignment model; Step S42: performing geometric shape adaptive generation according to the constraint assignment model to obtain a three-dimensional model to be verified; Step S43: Perform virtual performance index evaluation on the three-dimensional model to be verified to obtain a simulation performance report; Step S44: Perform model iterative optimization decision on the simulation performance report to obtain an adaptive three-dimensional model.

9. The intelligent three-dimensional model design method for spare parts according to claim 1, characterized in that: Step S5 includes the following steps: Step S51: monitoring the service status of the new model on the adaptive three-dimensional model to obtain the operation history of the new model; Step S52: Compare the actual performance data of the new model operation history with the behavior feature vector to obtain a performance difference report; Step S53: recording maintenance events and fault information of the adaptive three-dimensional model to obtain a new model maintenance file; Step S54: Calculate and evaluate the service performance index according to the performance difference report and the new model maintenance file to obtain a preliminary performance index set; Step S55: Perform design optimization strategy feedback on the preliminary performance indicator set to obtain service performance indicators.

10. An intelligent three-dimensional model design system for spare parts, characterized in that: For executing the intelligent three-dimensional model design method for spare parts as claimed in claim 1, the intelligent three-dimensional model design system for spare parts comprises: The dynamic feature extraction module is used to obtain multi-dimensional perception data of spare parts to obtain the original sensor data stream; to mine the associated feature parameters of the original sensor data stream to obtain the primary feature parameter set; to perform multi-dimensional feature fusion construction on the primary feature parameter set to obtain the behavior feature vector; The risk spectrum construction module is used to obtain the historical failure data of spare parts; train the multimodal failure prediction model based on the historical failure data of spare parts to obtain the multimodal failure prediction model; use the multimodal failure prediction model to deduce the future failure probability and accumulate the time series failure risk to obtain the failure probability distribution; The topological configuration evolution module is used to analyze the association between failure mechanism and design parameters of failure probability distribution to obtain a failure-parameter association table; to evaluate the sensitivity of design parameters on the failure-parameter association table to obtain parameter sensitivity evaluation data; to construct and optimize multi-objective constraints based on the parameter sensitivity evaluation data to obtain optimized design constraints; The parameterized kernel reconstruction module is used to perform constraint-driven parameter assignment according to the optimization design constraint to obtain a constraint assignment model; and to construct a geometric morphology adaptive three-dimensional model according to the constraint assignment model to obtain an adaptive three-dimensional model; The closed-loop performance evaluation module is used to monitor the service status of the new model of the adaptive three-dimensional model, generate a maintenance file, and obtain a new model maintenance file; and generate a service performance indicator according to the new model maintenance file to obtain a service performance indicator.

Images